Return and Improve Model - 2019 Edition

One of my pet projects for numerous year is back for another go-around. 2018 wasn't kind to the R&I model, and if I remember correctly, 2017 fudged a lot of the probabilities (I honestly can't remember, so here are the links if you are interested: 2017 and 2018). If you are unfamiliar with the tenants of the R&I model, the 2017 link will explain it in great detail. For the sake of time, I will be as brief as possible. The R&I model looks at the percentage of multiple statistical categories that a team returns from the previous year's team. It then forecasts the probability of the current year's team improving its tournament performance compared to the previous year's team based on the return-percentage. Since the model seems to have less and less applicability in this current era of one-and-done college basketball, this model and its probabilities have not been "qualified" or "scaled" based on any extenuating factors, such as critical match-ups, seed differentials, or era. Let's see what 2019 holds.

The Probabilities

Below is a chart of the return-counts and the improve-probabilities of all teams returning teams from 2001-2018.

For reference:

• xxR# is the number of teams that returned greater than xx-% of a particular stat.
• xxR&I# is the number of teams that returned greater than xx-% of a particular stat and increased their number of wins in the tournament (improved their tournament performance).
• xxR&I% is the Return and Improve Probability (xxR&I# / xxR#), if a team returns at least that percentage of that specific stat, then they have that particular probability of improving upon their previous year's tournament performance.

As one would logically expect, if a current team returns a high percentage of its statistics from last year's team, they should be a better team than last year's team and have a higher probability of improving upon last year's performance. Typically with the R&I model, I focus exclusively on Points Returned and Minutes Returned simply because points determines wins and losses in basketball and minutes are the best statistics to measure experience returning.

2019's R&I Candidates

The chart below shows the 23 teams in the 2019 tournament that also participated in the 2018 tournament. From a previous article (Bracket Profiles), we know a total of 30 teams returned between the two years. The R&I model only looks at the old power- Six conferences (ACC, B10, B12, SEC, P12, and Big East/American). As a result, this eliminates GONZ, BUFF, NEV, and the five small conference teams (12- thru 16-seeds). The R&I Model also looks at the play-in game participants even though the Tourney Profile model does not consider them, and this adds AZST to the R&I pool of participants (more on AZST later).

As you can see, we have a lot to deal with in 2019.

1. The reason I don't look at ranges below 40%-return is two-fold. First, if a team returns two key starters and no one else, that should be enough to put them in the 40% range, and I doubt that particular situation would tell us anything about returning-and-improving probabilities. Second, teams like UK and DUKE that build themselves on the one-and-done (OAD) approach will most likely influence the ranges below 40% based entirely on the quality of the freshman instead of the returning players. As for the remaining teams in this range -- TXTC, HALL, NOVA and KU -- the non-OAD teams in this range last year (UNC and UCLA) failed to improve. Considering the deep runs that three of these four need to make in order to improve, it's near-100% that these three fail and highly likely that all four teams fail to improve.
2. AZST is honestly another freebie. If they win, then they improve on last year's performance, and the probabilities for those ranges tick up without any bearing on the 2019 bracket. If they lose, then they fail to improve, and the probabilities for those ranges tick down without any bearing on the 2019 bracket.
3. Five teams above the 40% threshold have negative seed differentials, meaning their seeding is worse in 2019 than 2018, and those teams are AUB, FLA, CIN, OHST and PUR. In 2018, three out of four teams with this characteristic failed to improve (FLST defied this pattern).
4. Three teams above the 50% threshold have seed changes (SC) of four or greater, and they are KSU, VT, and FLST. In 2018, all four teams with this quality (MIST, CIN, XAV, and UVA) failed to improve. Two of these teams (FLST and KSU) need F4 runs in 2019 to improve upon their 2018 performance. Speaking of KSU, Dean Wade is counted as a non-returner in their numbers since he is likely to miss the opening round game to his lingering foot issue. If Wade was a returner, KSU would be near or above 90% in every statistical category.
5. Speaking of near 90% teams, the 2019 tournament provides us with two teams fitting this description: SYR and TENN. While 2018 had no such matches, 2017 did (WISC) and it presented us with a high-risk, high-reward opportunity. In 2017, WISC was an 8-seed (down from a 7-seed in 2016) that needed three wins to improve on 2016's performance, and this meant going through the tournament's top overall 1-seed NOVA. Well, WISC provided us the upset opportunity, but failed by a lay-up to win a third game (I had them in my personal bracket for three wins). SYR is in a very similar situation, and to make matters worse, if they defeat the fourth-overall 1-seed GONZ, they potentially might have to go through FLST, who is seeking a F4 run to improve upon their 2018 performance. This is the definition of a critical match-up.
1. This absolutely frightens me over GONZ's path to the F4. Being the fourth-overall 1-seed isn't a desirable position because only one (2008 KU) has won the NC since 2004 (when the 1-seeds were matched according to the S-Curve ranking).
2. I do like SYR's return percentages enough to pick them over BAY in their R64 match-up.
6. For any of the remaining teams, I do prefer a group-based approach. For example, teams that return 60-70% of their points and minutes have a 30-35% probability of improvement. In 2019, five teams fit this description (VT, FLST, AUB, UNC, UVA). Theoretically, two of these five teams should improve (40%), which would match the predictive probability (30-35%). 
1. The likeliest candidates would be UVA and VT because they only have to win one game. AUB and UNC only have to win two games to achieve improvement. Keep in mind, VT, FLST, and AUB were identified as potential fails in previous points (#3 and #4).
2. To take a contrarian approach, the 60-70% range has approximately similar probabilities as both the 50-60% range and the 40-50% range. Logic would lead you to believe that the probabilities for the 60-70% range should be better than, not equal to, the lower return ranges of 50-60% and 40-50%. To make a contrarian play based on this logic, you could predict all four of VT, AUB, UNC, and UVA since their requirements (either one or two wins) are not as steep as FLST's requirement (four wins). This doesn't mean FLST can't achieve their requirement, it just means the odds aren't favorable to them.
7. Oh yeah, there was a 2nd team (TENN) with near-90% returns, and it looks like I completely ignored them. I did not forget about them. I promise you it was intentional, and here is why. They only need two wins to improve upon their 2018 performance. It is certainly doable since their only R&I threat is 7-seed CIN, who was listed as a potential fail (Point #3). But I say, why stop after two. Here are four 1-win teams who didn't stop after two wins, and three of the four returned far less than 2019 TENN did. While some of my other models reject this proposition, I thought I would present the possibility.

Anyways, I hope you enjoyed this foray into the Return and Improve model. I should have my final predictions for the 2019 tournament released in about 24 hours.